PREDICTIVE VALUE OF DIGITAL RECTAL EXAMINATION FOR PROSTATE CANCER DETECTION IS MODIFIED BY OBESITY

Abstract

The American Cancer Society’s updated screening guidelines for prostate cancer (CaP) render digital rectal examination (DRE) optional. We investigated the impact of DRE on CaP detection among obese men. Data from 2794 men undergoing initial prostate biopsy at three centers were analyzed to assess CaP risk attributed to abnormal DRE across body-mass index (BMI) categories. Predictive accuracies of a combination of PSA, age, race, center, and biopsy year including or excluding DRE findings were compared by areas under the receiver-operating characteristics curves (AUC). In all cohorts, obese men were less likely to have abnormal DREs diagnosed than non-obese men. As BMI category increased, abnormal DREs became stronger predictors for overall CaP in individual (p-trends≤0.05) and combined (p-trend<0.001) cohorts, and for high-grade CaP in the Italian (p-trend=0.03) and combined (p-trend=0.03) cohorts. DRE inclusion improved the predictive accuracy for overall and high-grade CaP detection among all obese men (p≤0.032) but not normal-weight men (p≥0.198). DRE inclusion also near-significantly improved overall CaP detection in obese men with PSA<4ng/mL (p=0.081). In conclusion, the predictive value of DRE is dependent on obesity and is significantly higher among obese men than normal-weight men.

Introduction

Prostate cancer (CaP) and obesity remain closely-related healthcare problems affecting millions of men worldwide and incurring significant financial cost.^1,2 Obese men with CaP are at higher risk for adverse pathologic outcomes,³ biochemical recurrence,^4,5 and cancer-specific death^6–8 than non-obese men. One possible explanation has been sub-optimal screening due to lower PSA levels resulting from hemodilution.^9–11 Additionally, with adiposity impeding physical examinations,¹² obese men may face difficulties in undergoing a digital rectal examination (DRE) due to body habitus hindering prostate access.

Though screening remains controversial,¹³ most CaP cases today are detected via elevated PSA. This phenomenon has led to stage migration, with more men being stratified as low-risk with non-palpable disease.^14,15 Consequently, the role of DRE has been questioned^16–18 or even omitted in trials investigating CaP screening.^19,20 Recently, the American Cancer Society (ACS) revised its CaP screening guidelines, recommending DRE be an optional rather than necessary adjunct to PSA testing.²¹ However, some studies have suggested DRE remains relevant, particularly in detecting clinically-aggressive tumors.^22,23

To date, the predictive value of DRE in CaP detection specifically among obese men has never been examined. Our hypothesis was that an abnormal DRE might be a more significant predictor of CaP on biopsy in obese men than in non-obese men, even in those with low PSA values, because since DRE may be limited by adiposity in obese men, tumors large enough to be palpable may be more meaningful. Thus, we examined the impact of DRE findings on cancer detection as a function of obesity, using two multi-ethnic cohorts encompassing both tertiary-care and equal-access settings from the United States, and one European cohort from a university hospital in Italy.

Materials and Methods

Data were retrospectively abstracted after obtaining regulatory approval. A DRE was considered abnormal according to the physician’s discretion. All patients were referred for an initial prostate biopsy due to an abnormal DRE or PSA level. We excluded men with PSA≥20ng/mL due to high risk of having CaP and men who had had prior biopsies since an initial negative biopsy has been shown to reduce risk.²⁴ We calculated body-mass index (BMI) as weight divided by height squared (kg/m²).

Duke Prostate Center

We identified a multi-ethnic cohort of 1215 men with BMI information who underwent initial biopsy (2005–2009) at the Duke Prostate Center (DPC) in Durham, NC. Patients were excluded if they had PSA≥20ng/mL (n=26) or missing data for PSA (n=11), DRE (n=3), age (n=1), ethnicity (n=13), or biopsy outcome (n=33), resulting in a final population of 1145 men.

Durham Veterans Affairs Medical Center

We identified a multi-ethnic cohort of 1207 men with BMI information who underwent initial biopsy (2000–2009) at the Durham Veterans Affairs Medical Center (DVAMC). Patients were excluded if they had PSA≥20ng/mL (n=149) or missing data for PSA (n=1), DRE (n=118), or ethnicity (n=6), resulting in a final population of 933 men.

“La Sapienza” University of Rome

We identified 786 Italian men with BMI information who underwent initial biopsy (2002–2009) at “La Sapienza” University of Rome in Italy. Patients were excluded if they had PSA≥20ng/mL (n=65) or missing data for DRE (n=5), resulting in a final population of 716 men.

Statistical Analysis

We categorized BMI as normal-weight (<25kg/m²), overweight (25–29.9kg/m²), and obese (≥30kg/m²). Baseline variables were compared across BMI categories using chi-square/Fisher’s exact tests (categorical) and Kruskal-Wallis tests (continuous).

We calculated odds ratios for overall and high-grade (Gleason sum≥7 vs. negative biopsy/Gleason sum<7) CaP risk associated with an abnormal DRE using logistic regression adjusting for age (continuous), ethnicity (white, black, other), biopsy year (continuous), BMI (log-transformed continuous), and PSA (log-transformed continuous). Data for BMI and PSA were right-skewed requiring log-transformation. Due to significant interaction between DRE and BMI, we repeated the analyses stratified by BMI category and used tests for interaction to evaluate p-trend by incorporating cross-product terms for DRE and median BMI for each BMI category into regression models.

We assessed predictive accuracy of DRE in overall and high-grade CaP detection using receiver-operating characteristics (ROC) curves. We compared areas under the ROC curves (AUC) using non-parametric methods²⁵ in two fitted logistic regression models for each cohort: a reference model adjusted for age, year, ethnicity, log-transformed BMI, and log-transformed PSA, and a DRE model adjusted for the same covariates plus DRE. Analyses were repeated with stratification by BMI category. Sensitivity analyses were performed by restricting ROC curve analyses to men with PSA<4ng/mL and repeating analyses with PSA adjusted for hemodilution (7% and 16% increase above baseline PSA for overweight and obese men, respectively). Model calibrations were assessed by the Hosmer-Lemeshow statistic. Study cohorts were analyzed individually and in combination adjusting for center with a two-tailed alpha of 0.05.

Results

Overall, 45%, 48%, and 44% of men from the DPC, DVAMC, and Italian cohorts, respectively, had CaP diagnosed (Table 1). The DVAMC patients had the highest obesity rate (38%) and the Italians the lowest (20%). Obese men were less likely to be diagnosed with an abnormal DRE relative to normal- and overweight men in each cohort, though this association was significant only in DVAMC (p=0.005). Sub-analysis of distributions of number of total biopsy cores (≤6 vs. >6) did not find any significant differences across BMI categories in all three cohorts (data not shown).

TABLE 1.

Baseline demographic and clinicopathologic characteristics by patient cohort and BMI category.

Characteristic	Cohorts by BMI Category
	Duke Prostate Center (n=1145)				Durham VA (n=933)				La Sapienza University (n=716)
	<25	25–29.9	≥30	P1	<25	25–29.9	≥30	P1	<25	25–29.9	≥30	P1
No. pts (%)	283 (25)	486 (42)	376 (33)		208 (22)	373 (40)	352 (38)		214 (30)	358 (50)	144 (20)
Median (IQR):
Age at biopsy, y	63 (56–69)	63 (58–68)	61 (56–67)	0.051	63 (59–69)	63 (58–68)	63 (58–69)	0.409	67 (62–74)	67 (62–74)	67 (62–73)	0.912
Year of biopsy	2007 (2006–08)	2007 (2006–08)	2007 (2006–08)	0.075	2004 (2002–06)	2004 (2002–06)	2005 (2002–06)	0.148	2007 (2005–08)	2007 (2005–09)	2008 (2005–09)	0.008
PSA, ng/mL	5.0 (3.8–6.8)	4.9 (3.7–6.2)	5.0 (3.9–6.6)	0.263	5.9 (4.4–8.1)	5.6 (4.4–8.1)	5.6 (4.4–7.8)	0.941	5.9 (4.3–8.6)	6.5 (4.7–9.1)	6.7 (5.0–9.5)	0.200
n (%):
Ethnicity				<0.001				0.998				–
White	194 (69)	357 (73)	225 (60)		106 (51)	193 (52)	178 (51)		214 (100)	358 (100)	144 (100)
Black	75 (27)	109 (22)	240 (37)		100 (48)	177 (47)	171 (49)
Other	14 (5)	20 (4)	11 (3)		2 (1)	3 (1)	3 (1)
Abnormal DRE	106 (37)	181 (37)	123 (33)	0.311	69 (33)	94 (25)	73 (21)	0.005	61 (29)	80 (22)	30 (21)	0.156
Positive biopsy	126 (45)	210 (43)	183 (49)	0.266	118 (57)	163 (44)	168 (48)	0.010	104 (49)	147 (41)	64 (44)	0.212
Bx Gleason sum				0.126				0.904				0.063
<7	75 (60)	118 (56)	84 (46)		65 (55)	93 (57)	95 (57)		50 (49)	53 (37)	19 (30)
7 (3+4, 4+3)	33 (26)	65 (31)	67 (37)		42 (36)	60 (37)	61 (36)		41 (40)	68 (48)	29 (46)
>7	18 (14)	27 (13)	32 (17)		11 (9)	10 (6)	12 (7)		11 (11)	22 (15)	15 (24)

NOTE: Values may not sum to the total study population for each database due to rounding or unavailability of data.

Abbreviations: BMI, body-mass index; PSA, prostate-specific antigen; DRE, digital rectal exam; IQR, inter-quartile range.

¹p-values by Kruskal-Wallis test for continuous variables and chi-square/Fishers exact test for categorical variables.

On multivariable analysis adjusted for BMI as a continuous variable, an abnormal DRE portended nearly twice the odds of any CaP compared to normal DRE in all three cohorts (p<0.001; Table 2). When stratified by BMI category, overweight and obese men with an abnormal DRE in all three cohorts had nearly two-fold to greater than six-fold higher odds of positive biopsy compared to a normal DRE (p≤0.003). Tests for trend of risk across BMI categories were significant: as BMI category increased, abnormal DREs became stronger predictors for CaP in all individual (p-trends≤0.05) and combined (p-trend<0.001) cohorts.

TABLE 2.

Risk for overall CaP detection with abnormal DRE compared to normal DRE.

		By BMI Category
Study Cohort	Overall	<25	25–29.9	≥30	p-trend2
Duke Prostate Center					0.03
OR1	1.87	1.22	1.89	2.93
95% CI	1.43 – 2.45	0.73 – 2.04	1.24 – 2.87	1.75 – 4.92
P	<0.001	0.456	0.003	<0.001
Durham VA					0.04
OR1	2.59	2.21	2.32	3.88
95% CI	1.86 – 3.60	1.08 – 4.55	1.40 – 3.84	2.15 – 7.02
P	<0.001	0.031	0.001	<0.001
La Sapienza University					0.05
OR1	2.82	2.13	2.60	6.66
95% CI	1.94 – 4.11	1.10 – 4.12	1.51 – 4.47	2.45 – 18.07
P	<0.001	0.025	0.001	<0.001
COMBINED3					<0.001
OR1	2.22	1.57	2.11	3.43
95% CI	1.86 – 2.66	1.11 – 2.21	1.61 – 2.77	2.42 – 4.86
P	<0.001	0.011	<0.001	<0.001

Abbreviations: CaP, prostate cancer; BMI, body-mass index; DRE, digital rectal exam; OR, odds ratio; CI, confidence interval.

¹For overall model with BMI as continuous variable, DRE results adjusted for log BMI, log PSA, age at biopsy, year of biopsy, and ethnicity. For models stratified by BMI category, DRE results adjusted for log PSA, age at biopsy, year of biopsy, and ethnicity. All models constructed using multivariable logistic regression.

²Test for trend of increasing odds ratios across BMI categories.

³Combined analyses additionally adjusted for study center.

Patients with abnormal DREs also had more than twice the risk of having high-grade CaP adjusting for BMI as a continuous covariate (p<0.001; Table 3). This increased risk associated with abnormal DRE was again noted in all BMI categories in all cohorts except normal-weight men in DPC (p=0.062). Although the odds for high-grade CaP detection portended by an abnormal DRE were higher among obese relative to normal-weight men in all three cohorts, trends of increasing risk across increasing BMI categories reached significance only in the Italian (p-trend=0.03) and combined (p-trend=0.03) cohorts.

TABLE 3.

Risk for Gleason Sum ≥7 CaP detection¹ with abnormal DRE compared to normal DRE.

		By BMI Category
Study Cohort	Overall	<25	25–29.9	≥30	p-trend3
Duke Prostate Center					0.41
OR2	2.41	1.85	2.53	2.74
95% CI	1.76 – 3.30	0.97 – 3.51	1.52 – 4.21	1.61 – 4.66
P	<0.001	0.062	<0.001	<0.001
Durham VA					0.26
OR2	3.51	2.84	3.32	4.46
95% CI	2.44 – 5.05	1.36 – 5.92	1.84 – 5.96	2.41 – 8.24
P	<0.001	0.006	<0.001	<0.001
La Sapienza University					0.03
OR2	2.82	2.08	3.61	7.29
95% CI	1.94 – 4.11	1.01 – 4.27	2.02 – 6.46	2.84 – 18.72
P	<0.001	0.046	<0.001	<0.001
COMBINED4					0.03
OR2	2.76	2.03	2.78	3.60
95% CI	2.26 – 3.37	1.37 – 3.01	2.03 – 3.80	2.51 – 5.16
P	<0.001	<0.001	<0.001	<0.001

Abbreviations: CaP, prostate cancer; BMI, body-mass index; DRE, digital rectal exam; OR, odds ratio; CI, confidence interval.

¹In reference to men who either had Gleason Sum <7 tumors or a negative biopsy.

²For overall model with BMI as continuous variable, DRE results adjusted for log BMI, log PSA, age at biopsy, year of biopsy, and ethnicity. For models stratified by BMI category, DRE results adjusted for log PSA, age at biopsy, year of biopsy, and ethnicity. All models constructed using multivariable logistic regression.

³Test for trend of increasing odds ratios across BMI categories.

⁴Combined analyses additionally adjusted for study center.

Full non-stratified regression models in all cohorts that included DRE showed significantly higher predictive accuracies for overall CaP detection than models that excluded DRE (p≤0.016; Table 4). Upon stratification by BMI category, obese men displayed significant improvements in AUC values by inclusion of DRE (p≤0.019), while normal-weight men did not (p≥0.202). These findings held true in the combined analysis (obese men: AUC 0.67 vs. 0.61, p<0.001; normal-weight men: AUC 0.68 vs. 0.67, p=0.365), indicating that the additional predictive value of DRE in CaP detection was dependent on BMI category and was greatest in obese men and least in normal-weight men.

TABLE 4.

Predictive accuracies for overall CaP detection in regression models with/without DRE findings.

Study Cohort	Overall	By BMI Category
		<25	25–29.9	≥30
Duke Prostate Center
Reference Model1  AUC (95% CI)	0.63 (0.60–0.66)	0.64 (0.58–0.71)	0.62 (0.57–0.67)	0.68 (0.62–0.73)
DRE Model2  AUC (95% CI)	0.65 (0.62–0.69)	0.65 (0.58–0.71)	0.64 (0.59–0.69)	0.72 (0.67–0.77)
P3	0.016	0.475	0.302	0.019
Durham VA
Reference Model1  AUC (95% CI)	0.61 (0.57–0.64)	0.73 (0.66–0.79)	0.61 (0.55–0.67)	0.56 (0.50–0.62)
DRE Model2  AUC (95% CI)	0.64 (0.61–0.68)	0.73 (0.67–0.80)	0.64 (0.59–0.70)	0.65 (0.59–0.70)
P3	0.009	0.664	0.101	0.003
La Sapienza University
Reference Model1  AUC (95% CI)	0.66 (0.62–0.70)	0.69 (0.62–0.76)	0.67 (0.62–0.73)	0.62 (0.52–0.71)
DRE Model2  AUC (95% CI)	0.70 (0.66–0.74)	0.71 (0.64–0.78)	0.70 (0.65–0.76)	0.70 (0.61–0.79)
P3	0.003	0.202	0.050	0.013
COMBINED4
Reference Model1  AUC (95% CI)	0.63 (0.60–0.65)	0.67 (0.63–0.71)	0.62 (0.59–0.66)	0.61 (0.57–0.65)
DRE Model2  AUC (95% CI)	0.65 (0.63–0.67)	0.68 (0.64–0.72)	0.65 (0.62–0.68)	0.67 (0.63–0.71)
P3	<0.001	0.365	0.016	<0.001

Abbreviations: CaP, prostate cancer; BMI=body-mass index; DRE, digital rectal exam; AUC=area under receiver-operating characteristics curve; CI=confidence interval.

¹Overall model with BMI as continuous variable adjusted for log BMI, log PSA, age at biopsy, year of biopsy, and ethnicity. Models stratified by BMI category adjusted for log PSA, age at biopsy, year of biopsy, and ethnicity. All models constructed using multivariable logistic regression.

²Adjusted for DRE results, in addition to covariates for Reference Models using multivariable logistic regression.

³Test of equality for AUCs between Reference and DRE Models.

⁴Combined analyses additionally adjusted for study center.

Similar results were seen for high-grade CaP detection, with models including DRE having better accuracy than models excluding DRE in the full non-stratified analysis (p≤0.008; Table 5). Stratification by BMI category again showed improvements in accuracies by DRE inclusion for obese men in all individual (p≤0.032) and combined (p<0.001) cohorts, but not for normal-weight men (p≥0.198). For overweight men, accuracies were higher in favor of DRE inclusion in the Italian (p=0.020) and combined (p=0.005) cohorts.

TABLE 5.

Predictive accuracies for Gleason Sum ≥7 CaP detection¹ in regression models with/without DRE findings.

Study Cohort	Overall	By BMI Category
		<25	25–29.9	≥30
Duke Prostate Center
Reference Model2  AUC (95% CI)	0.68 (0.64–0.72)	0.69 (0.62–0.76)	0.69 (0.63–0.75)	0.69 (0.63–0.75)
DRE Model3  AUC (95% CI)	0.71 (0.67–0.75)	0.71 (0.64–0.78)	0.72 (0.66–0.78)	0.73 (0.67–0.79)
P4	0.008	0.201	0.144	0.032
Durham VA
Reference Model2  AUC (95% CI)	0.66 (0.61–0.70)	0.75 (0.68–0.83)	0.64 (0.57–0.71)	0.63 (0.56–0.70)
DRE Model3  AUC (95% CI)	0.71 (0.67–0.76)	0.78 (0.70–0.85)	0.70 (0.63–0.76)	0.71 (0.65–0.78)
P4	<0.001	0.304	0.066	0.018
La Sapienza University
Reference Model2  AUC (95% CI)	0.69 (0.65–0.73)	0.69 (0.61–0.77)	0.70 (0.64–0.76)	0.67 (0.58–0.76)
DRE Model3  AUC (95% CI)	0.74 (0.69–0.78)	0.71 (0.63–0.79)	0.75 (0.69–0.80)	0.77 (0.69–0.86)
P4	0.004	0.457	0.020	0.008
COMBINED5
Reference Model2  AUC (95% CI)	0.67 (0.64–0.69)	0.69 (0.65–0.74)	0.67 (0.64–0.71)	0.66 (0.62–0.70)
DRE Model3  AUC (95% CI)	0.70 (0.68–0.73)	0.71 (0.66–0.75)	0.71 (0.67–0.74)	0.72 (0.68–0.76)
P4	<0.001	0.198	0.005	<0.001

Abbreviations: CaP, prostate cancer; BMI=body-mass index; DRE, digital rectal exam; AUC=area under receiver-operating characteristics curve; CI=confidence interval.

¹In reference to men who either had Gleason Sum <7 tumors or a negative biopsy.

²Overall model with BMI as continuous variable adjusted for log BMI, log PSA, age at biopsy, year of biopsy, and ethnicity. Models stratified by BMI category adjusted for log PSA, age at biopsy, year of biopsy, and ethnicity. All models constructed using multivariable logistic regression.

³Adjusted for DRE results, in addition to covariates for Reference Models using multivariable logistic regression.

⁴Test of equality for AUCs between Reference and DRE Models.

⁵Combined analyses additionally adjusted for study center.

We specifically tested the value of DRE in men with PSA<4ng/mL. As each cohort contained modest numbers of men with PSA<4ng/mL, they were combined for analysis. Of 581 total men with PSA<4ng/mL, 198 had CaP and 73 had high-grade CaP detected on biopsy (Table 6). DRE inclusion improved the accuracy for overall CaP detection that neared statistical significance among obese men (AUC 0.75 vs. 0.70, p=0.081) but not among normal- or overweight men (p≥0.243). There was no added benefit of DRE inclusion for high-grade CaP detection in obese men with PSA<4ng/mL (AUC 0.74 vs. 0.72, p=0.591) though only 26 such men had high-grade CaP. Repeating all analyses with PSA values adjusted for hemodilution resulted in identical findings (data not shown). All models created had non-significant Hosmer-Lemeshow statistics, indicating good calibration.

TABLE 6.

Predictive accuracies for overall and Gleason Sum ≥7 CaP detection¹ in combined regression models with/without DRE findings in patients with PSA<4ng/mL.

COMBINED	Overall	By BMI Category
		<25	25–29.9	≥30
No. pts (%) with PSA<4ng/mL	581	159 (27)	245 (42)	177 (30)
Overall CaP, n (%)	198	52 (27)	80 (41)	65 (33)
Reference Model2  AUC (95% CI)	0.67 (0.63–0.72)	0.68 (0.59–0.77)	0.69 (0.61–0.76)	0.70 (0.63–0.78)
DRE Model3  AUC (95% CI)	0.68 (0.64–0.73)	0.68 (0.59–0.77)	0.70 (0.63–0.77)	0.75 (0.67–0.82)
P4	0.170	0.705	0.243	0.081
High-grade CaP1, n (%)	73	17 (23)	30 (41)	26 (36)
Reference Model2  AUC (95% CI)	0.69 (0.63–0.75)	0.79 (0.69–0.89)	0.68 (0.58–0.79)	0.72 (0.64–0.81)
DRE Model3  AUC (95% CI)	0.69 (0.63–0.75)	0.79 (0.69–0.89)	0.71 (0.61–0.80)	0.74 (0.64–0.83)
P4	0.531	0.900	0.444	0.591

NOTE: Percentages may not sum to the total due to rounding.

Abbreviations: CaP, prostate cancer; BMI=body-mass index; DRE, digital rectal exam; AUC=area under receiver-operating characteristics curve; CI=confidence interval.

¹In reference to men who either had Gleason Sum <7 tumors or a negative biopsy.

²Overall model with BMI as continuous variable adjusted for log BMI, log PSA, age at biopsy, year of biopsy, ethnicity, and study center. Models stratified by BMI category adjusted for log PSA, age at biopsy, year of biopsy, ethnicity, and study center. All models constructed using multivariable logistic regression.

³Adjusted for DRE results, in addition to covariates for Reference Models using multivariable logistic regression.

⁴Test of equality for AUCs between Reference and DRE Models.

Discussion

Though it may cause mild discomfort, DRE provides a safe and minimally-invasive technique for prostate evaluation among men with varying BMI. We noted the risk of CaP on initial biopsy with an abnormal DRE is nearly two- to three-fold higher in obese versus normal-weight patients. Furthermore, we showed that including DRE with PSA testing gives a significantly higher predictive accuracy than without DRE for CaP detection, but only among obese men and not normal-weight men. Most importantly, our findings, though potentially underpowered, showed trends toward significance even in men with PSA<4ng/mL, who may potentially benefit the most from DRE. These results suggest that while an abnormal DRE may be an important predictor for CaP in normal-weight men, it favors CaP detection to an even greater degree in obese men.

The revised ACS guidelines recommend screening with PSA testing with or without DRE starting at age 50 for asymptomatic men of average risk with life-expectancies greater than 10 years.²¹ Notably, these guidelines differ from those of the American Urological Association²⁶ and European Association of Urology²⁷ that include DRE in screening. Most contemporary tumors are detected solely by PSA screening, which has a greater detection rate compared to DRE alone.²⁸ Consequently, the use of DRE for screening has received criticism due to low sensitivity.^16–18 The rationale behind the ACS recommendations²¹ stems from results of the European Randomized study of Screening for Prostate Cancer (ERSPC) and the Prostate, Lung, Colorectal, and Ovarian (PLCO) cancer screening trial. The ERSPC screened the majority of men with PSA without DRE and found a mortality reduction,²⁹ whereas the PLCO study offered both PSA and DRE, but did not demonstrate any benefit.³⁰ Additionally, sub-group analyses of men who underwent DRE in ERSPC found only 17% of CaP were solely detected by DRE.¹⁶ Thus, despite evidence in PLCO of poor biopsy compliance after positive DRE screen³¹ and widespread contamination in the control arm,³² the ACS concluded randomized trial evidence did not support DRE as a CaP screening test.²¹

Obese men, however, constitute a select group in whom we have shown an abnormal DRE remains highly significant for predicting CaP, and even possibly in those with PSA<4ng/mL. Moreover, we found the value of DRE as a CaP predictor to be consistent across multiple centers, ethnicities, and practice settings. Indeed, the significance of detecting CaP among obese men cannot be overstated. Obese men are at higher risk for post-prostatectomy recurrence^4,5 and disease-specific death.^6–8 Prior studies suggested current screening practices may be biased against obese men due to hemodiluted PSA secondary to increased plasma volume^9,10 or to larger-sized prostates hiding tumors from DREs and biopsies.^33,34 A recent analysis of three national surveys showed although obese men were no less likely to undergo PSA testing than non-obese men, they had lower PSA-driven biopsy rates.³⁵ Additionally, one study reported obese men were less likely to undergo DRE as part of CaP screening.³⁶ As such, should DRE be made optional, obese men in particular may face even greater disparity in CaP screening than their non-obese peers, while they may in fact benefit the most from DRE.

To our knowledge, we are the first to compare the significance of DRE in detecting CaP across BMI categories. As DRE is operator-dependent, one crucial factor is feasibility in performing a thorough DRE in any man. It is difficult to find DRE abnormalities if the entire prostate cannot be palpated. One possible explanation for why DRE in our study was more predictive among obese men is that DRE may be limited by excess adiposity, meaning tumors that must be large enough to be palpable probably are more significant. Indeed, obese patients have been noted to have larger tumors and higher-grade disease at time of prostatectomy.³ Thus, while obese men in our study were less likely to have an abnormal DRE diagnosed, such a finding predicted a significantly greater chance of a positive biopsy than an abnormal DRE in non-obese men.

Importantly, we showed DRE adds significant benefit to PSA accuracy, but only for obese and some overweight men. Normal-weight men showed no improvement in cancer detection when DRE findings were included with PSA. One study found improved diagnostic performance for combination PSA and DRE for CaP detection versus either alone, but did not stratify by BMI.²⁸ Our findings suggest one potential hypothesis to explain the discrepancy in added benefit of DRE in early cancer detection in previous studies:^16–18 Europeans are less obese than Americans. Indeed, the Italian cohort had barely half the proportion of obese men as the DVAMC patients. However, among these obese Italian men, DRE inclusion to PSA still significantly increased the predictive accuracy of both overall and high-grade CaP detection. Moreover, in our sensitivity analysis of men with PSA<4ng/mL, we found that among obese men in this subpopulation, DRE inclusion increased predictive accuracy for overall CaP detection from 0.70 to 0.75 with a trend nearing significance despite the limited number of patients (n=65). Thus, obese men may potentially reap the greatest benefit in early CaP detection from DRE, though the danger of over-diagnosing indolent disease exists.

Some studies emphasized the importance of an abnormal DRE in predicting high-grade CaP.^23,28 Our adjusted models found this association for Gleason sum≥7 tumors in nearly all the patients, with a significant trend of increasing risk across increasing BMI categories in the combined analysis. Notably, while DRE inclusion improved predictive accuracy in high-grade cancer detection adjusting for BMI as a continuous variable, we again found the added predictive value of DRE was BMI-dependent. All obese and most overweight men showed significant improvement in predictive accuracy with DRE inclusion, while normal-weight men did not. This is important because higher-grade disease may be part of the reason for poorer outcomes among obese men.³ Although we did not note a significant improvement in predicting high-grade CaP with DRE inclusion for obese men with PSA<4ng/mL, we hypothesize we were underpowered (n=26).

A key strength of our study was the use of three large cohorts from two countries with contrasting obesity rates and contrasting referral patterns. While our findings may be thought to be applicable only to Americans due to higher prevalence of obesity in the US, corroboration of these results among Europeans support validity and generalizability of these findings. Interestingly, we found that an abnormal DRE has greater significance in obese Italian men versus obese American men from the South, which in the US is notorious for its high obesity rates.³⁷ Furthermore, our results were not changed after incorporation of BMI-adjusted PSA values, indicating DRE findings improved predictive accuracy for CaP detection despite possible effects of PSA hemodilution. Lastly, we included two multi-ethnic cohorts to help adjust for possible racial differences in tumor biology, although previous work showed the relationship between obesity and adverse CaP outcomes is likely independent of race.⁵

Our study has limitations. First, our study populations are not screening cohorts, but rather referred men undergoing initial biopsy, which may limit generalizability of our results. However, biopsy cohorts have the advantage of cancer status ascertainment. Secondly, we cannot fully differentiate non-thorough incomplete DREs—a common obstacle among obese men—from thoroughly-performed negative DREs. A DRE, however, would still have been attempted in an incomplete DRE, and not every obese man will have an incomplete DRE. Thirdly, we cannot control for differences in CaP screening histories prior to biopsy, which may affect DRE test performance. CaP diagnosis, however, still requires biopsy, which we did control for by excluding men with prior biopsies. Fourthly, the number of men with PSA<4ng/mL in our referred patient populations was limited, requiring validation in larger cohorts. Lastly, our study was retrospective. However, both PLCO and ERSPC did not analyze men by BMI category, which was the primary strategy of our study. Ultimately, whether screening with DRE in obese men leads to increased high-grade CaP detection and ultimately survival while minimizing over-detection remains to be determined.

Conclusions

We found the predictive value of an abnormal DRE in cancer detection is significantly modified by a patient’s BMI. Specifically, an abnormal DRE has nearly two- to three-folds greater significance in predicting CaP risk among obese compared to normal-weight men undergoing initial biopsy. Furthermore, incorporating DRE with PSA improved predictive accuracy for CaP detection in obese but not normal-weight men, with a trend towards significance even among those with PSA<4ng/mL. These findings should, at the very least, raise concern regarding generalizability of the newest ACS recommendations that render DRE optional to all men regardless of BMI. With increasing global obesity rates and excess weight being closely linked with screening biases and increased CaP death, all available tools to optimize CaP detection for timely intervention among obese men should be utilized. As such, it may be advisable for DRE to remain an essential component of CaP evaluation, particularly in obese men.